Facebook

Login

Support Sailonline

If you haven't already - join the SAILONLINE YACHT CLUB!

Please also consider making a donation - all amounts are greatly appreciated!

Board » Technical Discussion » Performance loss

Page: First Previous 3 4 5 6 7 8 9 Next Last

Posted by
kroppyer


Users avatar

Posts
347
Registered at
2010-02-05 15:25:55
2013-10-15 12:48:46 - Re: Performance loss
I've run some more tests, and sometimes the models predict a value that is so close to the actual one that it's scary, I've seen a difference of 0.000000004 (I might have missed a zero).

Looking at my results, I think we have (almost) unravelled the performance recovery model:
new_perf = min(100, old_perf + delta_time*3/(20*abs(boatspeed)))
> new_perf and old_perf in %
> boatspeed is actual boatspeed (not boatspeed when perf would be 100%) from previous "server jump" in knots
> delta_time: time since previous "server jump", in seconds
> abs(x) makes x positive
> min(x, y) returns the smallest value of x and y.
This formula basically translates to
"Your performance will increase with 9/boatspeed% per minute, until you reach 100%"

The performance loss model that matches the actual values the best, is really close, but no perfect match:
if(tack or gybe and old_perf >= 93)
new_perf = old_perf * (1 - abs(boatspeed)/200)
else if(normal course change and old_perf >= 93)
new_perf = old_perf * (1 - dTWA/25)
end if
> new_perf and old_perf in %
> boatspeed in knots
> dTWA is the positive difference between TWA before and after the course change, in radians (1 radian = 180/pi degrees)

I think the last thing to unravel is what boatspeed to use in the calculation for performance loss while taking/gybing.
> new boatspeed if perf would be 100%
> old boatspeed if perf would be 100%
> new boatspeed if perf would be the the same as before the tack
> old actual boatspeed (with performance the same as before the tack)

And there's another thing to do with performance loss that I we haven't tried yet: bbq'ing, I'm not sure why we would want to know the exact performance loss after a bbq, but why not?

Attached:

> Image with actual values, and values as predicted by 4 different models:
- model 1: use current boatspeed for tack/gybe-perf.loss-calculation
- model 2: use max boatspeed (when perf would be 100%) for tack/gybe-perf.loss-calculation
- model 3: like model 1, but performance loss is directly subtracted from performance (instead of multiplying with the (1 - loss) thing.
- model 4: like model 2, but same substraction method as in model 3

model 1 fits best.

> Raw data, csv file, can be opened with excel
Attachments
Posted by
javakeda


Users avatar

Posts
73
Registered at
2013-03-01 03:18:22
2013-10-15 14:08:25 - Re: Performance loss
Kroppyer wrote:

>>>>>
The performance loss model that matches the actual values the best, is really close, but no perfect match:
if(tack or gybe and old_perf >= 93)
new_perf = old_perf * (1 - abs(boatspeed)/200)
else if(normal course change and old_perf >= 93)
new_perf = old_perf * (1 - dTWA/25)
end if

>>>>>

So, just to complete the logic [and make sure that I am following you] ...
If the code falls through the IFs above, we are left with:

new_perf = min(100, old_perf + delta_time*3/(20*abs(boatspeed)))

And that implements the 93% rule?

Can you determine from the data whether the "recovery" code executes before, after, or instead of the "course change" code?

Thank you for your work on this.

Posted by
Go4iT


Users avatar

Posts
130
Registered at
2011-03-14 23:58:15
2013-10-15 15:03:51 - Re: Performance loss
Looks like it is about time for the SOL developers to change the performance loss parameters! hehehe
Posted by
Rod


Users avatar

Posts
392
Registered at
2010-06-24 22:30:48
2013-10-15 16:44:33 - Re: Performance loss
Why? It wasn't supposed to be a secret, was it? However, it does seem somewhat complicated, unless it was derived from actual data gathered from IRL boats. I would question, however the perf loss from a gybe,, which seems too great, UNLESS it involves dropping and re-hoisting a spinnacker. And a mere change of course should not involve a perf loss of any great amount. The rudder is still a good brake, however!
I am going to attempt to reduce the math to two ready-reference graphs, one showing perf loss at any angle and wind-speed, the other showing expected recovery and time to add to any DC to compensate for the loss in distance travelled.
If it breaks, it's not strong enough--if it doesn't, it's too heavy.
Posted by
A2


Users avatar

Posts
347
Registered at
2009-10-26 00:39:42
2013-10-15 17:16:24 - Re: Performance loss
Why Rod ? Why Indeed !

The Rationale for this hindrance has been expounded on in the previously linked thread.

Wrong Speed VMG

Posted by
Jakob Kuttenkeuler
2009-02-12

"We introduced this performance loss not to punish normal sailing maneouvers but to prevent boats to tack e.g. once every minute during hours just to ride on a tws that optimizes vmg. The performance loss implementation succeeded in preventing that behavior and is as stated above not a problem for normal maneouvers."

But please don't let me stop anyone from actually going to the source material and reading it for themselves as there are even more useful tidbits to be gleaned should one choose to do so.

There is merit in spending an equal amount of time reading these Forums as is devoted to contributing in them.

None so blind
Posted by
javakeda


Users avatar

Posts
73
Registered at
2013-03-01 03:18:22
2013-10-15 19:52:44 - Re: Performance loss
A2R: Thanks for the link to that thread.

To think that the PERF formulas have been sitting on the forum since 2009!
Ya think we need a search function for this puppy?

Or, as you suggested, ya think we need to RTFF?

(Note: Those who know me as an avid fan of MAD Magazine will recognize that the first 'F' in the acronym above stands for 'Furshlugginer'.)
Posted by
A2


Users avatar

Posts
347
Registered at
2009-10-26 00:39:42
2013-10-15 20:18:38 - Re: Performance loss
NSA recommends using Google for this purpose

site:http://www.sailonline.org/board/ "performance loss"

yeilded only only 9 hits when I went looking for it. Today there are 11. Took about all of 20 minutes to find the good stuff. Time well spent considering.
None so blind
Posted by
kroppyer


Users avatar

Posts
347
Registered at
2010-02-05 15:25:55
2013-10-15 20:28:49 - Re: Performance loss
Performance Recovery
new_perf = min(100, old_perf + delta_time*3/(20*abs(boatspeed)))
> new_perf and old_perf in %
> boatspeed is actual boatspeed (not boatspeed when performance would be 100%) from previous performance calculation. Speed in knots
> delta_time: time since previous performance calculation, in seconds
> abs(x) makes x positive
> min(x, y) returns the smallest value of x and y.
This formula basically translates to
"Your performance will increase with 9/boatspeed% per minute, until you reach 100%"

Performance loss
if(tack or gybe and old_perf >= 93)
new_perf = old_perf * (1 - abs(boatspeed)/200)
else if(normal course change and old_perf >= 93)
new_perf = old_perf * (1 - dTWA/25)
end if
> new_perf and old_perf in %
> boatspeed is the boat speed after the tack, calculated using the old performance. Speed in knots
> dTWA is the positive difference between TWA before and after the course change, in radians (1 radian = 180/pi degrees)

I believe performance loss only kick in when a command is executed, or when the boat tacks or gybes (when you are sailing cc trough a windshift, your performance drops when you gybe/tack)

I think performance recovery is also active in the "server jump" that performance loss is calculated. But the command to change course can be at the start of the server jump, or at the end, or somewhere in between. I'm not sure what boatspeeds are used how in the recovery formula in the same jump performance loss is calculated.

I think we have reach the limit now, the server jumps limit us from being more accurate, and I think it's not really of any use to be able to predict performance more accurate.

I will, when I caught up with other things I should've done, implement this performance model in my program spinnacer, so that calculations with performance will be easier to do.

---

It might be interesting to think of tricks with this model. hmm did his trick with the gybe, but I'm sure there's more:
- a 100.3 degree course change will cause 7% performance loss. So when you need to do a sharper turn (without gybing/tacking) you can turn 105 degrees first (perf<93%), and directly after that the rest of the turn (perf does not drop anymore than those 93%)
- Performance loss is smaller when your performance is not yet at 100%, but you're faster when you wait for your performance to reach 100% (sailing two times 5 minutes with average performance 98.5% is faster than sailing 10 minutes with an average performance of 97.2%). But what if your perf is lower than 93%, it's of course better to do it at that moment, as it won't do anything with your performance. And then there's an area in between, where part of the perf. loss could be "absorbed" by the 93% rule.
- Tacking, and then turning to the right course is extra performance loss, if you tack to the right course in one go, it faster, unless boat speeds differ much.. It's getting complicated..
- It in light winds, it might be better to do two tacks/gybes instead of one course change. Since tacking/gybing doesn't result in a big drop in performance when sailing slow, while a course change always has the same performance drop.
- etc.. etc..

I think we have unravelled the model and calculations, now comes the hard part: how to use it in our advantage :) Anyway, with this we can at least do better calculations for DCs/routing
Posted by
Go4iT


Users avatar

Posts
130
Registered at
2011-03-14 23:58:15
2013-10-15 23:15:49 - Re: Performance loss
A2R said @ 2013-10-15 20:18:38:
NSA recommends using Google for this purpose

site:http://www.sailonline.org/board/ "performance loss"

yeilded only only 9 hits when I went looking for it. Today there are 11. Took about all of 20 minutes to find the good stuff. Time well spent considering.
Tried that as a search engine in Google and got '404'. What am I doing wrong?

--- Last Edited by Go4iT at 2013-10-15 23:16:35 ---
Posted by
A2


Users avatar

Posts
347
Registered at
2009-10-26 00:39:42
2013-10-15 23:29:08 - Re: Performance loss
Go to Google;

In the Search bar

site:

followed without any space
by the url of this board

then your search terms after that

You can also go to the link at the bottom of the Google search page and click the link that says "Advanced Search"

There are boxes there to do the same thing, so when you hit search look at how Google has arranged the same business in the Search bar for you.



--- Last Edited by A2R at 2013-10-16 00:12:13 ---
None so blind
Attachments

Page: First Previous 3 4 5 6 7 8 9 Next Last

Please login to post a reply.

Races

Next Race: 00d 00h 00m

Current Races:

Cross the Convergence 2026 - Cocos to Sagar Island


Kapil Muni, incarnation of Lord Vishnu, founder of the Samkhya school of philosophy, disturbed from his meditation, in his wrath reduced the 60,000 sons of Sagar to ashes, as they dug the earth to retrieve the sacrificial horse that Lord Indra, the king of the Gods, stole off Sagar, thus creating the Sagara (ocean). Aeons later, Bhagiratha, a descendant of Sagara, asked Ganga to cleanse the ashes of the 60,000 princes, and her resultant torrential fall from heaven created the great river and where it met the ocean, the princes’ souls were washed and raised to heaven and Gangasagar (Sagar Island) was formed. So, let us join the pilgrimages to Kapi Muni’s Temple (see above) and race 2400nm from Cocos Island deep in the Sagara thence, in rather rapid Rapido 60s.
Race #1987
INFOby brainaid.de
Rapido 60PARTICULARS
WX Updates:
0430 / 1030 / 1630 / 2230
Ranking: CCZ - SYC
Race starts: Sep 11th 18:00 Registration Open!
▶ Flash
GO TO RACE

The Vineyard Race 2026


For the 16th consecutive year, our SOLers are invited to tackle this iconic 250nm overnighter. The classic route charges out through the sound, leaves Buzzards Bay Light to starboard, rounds Block Island to starboard, and heads home.

This year's weapon of choice is the Beneteau First 47.7. A proven performer since its 2012 SOL debut, this Farr-designed cruiser/racer excels at leading real and virtual handicap fleets. Success will demand sharp tactical navigation, so watch the polars closely to conquer the sound's shifting winds!
Race #2056
INFO by brainaid.de
First 47.7 PARTICULARS
WX Updates:
0430 / 1030 / 1630 / 2230
Ranking: SUPBUD - SYC
Race starts: Sep 04th 17:05 Registration Closed
▶ Flash
GO TO RACE

Shearwater - Rio de la Plata to Irish Sea 2026

Welcome to the fourth leg of our MIG series. This time we're racing the Sooty Shearwater on their migration from the breeding grounds in South America to the northern Atlantic. This is also part of our Ocean Championship, and as such the third race of OCQ3 2026. This race will be sailed in our Gunboat 90. Race #2055
INFO by brainaid.de
GunBoat 90 PARTICULARS
WX Updates:
0430 / 1030 / 1630 / 2230
Ranking:
OCQ3 - OCCH - SYC - MIG
Race starts: Sep 02nd 11:00 Registration Closed
▶ Flash
GO TO RACE

Go to race archive

SYC Ranking

  1. Sailonline Yacht Club Member CriticalHippo
  2. Sailonline Yacht Club Member KaSToR
  3. Sailonline Yacht Club Member WRmirekd
  4. Sailonline Yacht Club Member rafa
  5. Sailonline Yacht Club Member Panpyc
  6. Sailonline Yacht Club Member cdhinman
  7. Sailonline Yacht Club Member rthinman
  8. Sailonline Yacht Club Member BRENTGRAY
  9. Sailonline Yacht Club Member Sax747
  10. Sailonline Yacht Club Member vida

View full list

Series

Mobile Client

SYC members have the benefit of access to our mobile/lightweight web client!

The mobile client